Combinatorial libraries of synthetic and natural products are important sources of molecular information for the development of pharmacologic agents. Linear peptide libraries, containing known and random peptide sequences, are particularly good sources of new and novel compounds for drug development because of the diversity of structures which can be generated. Drawbacks to linear peptide libraries include the following: (1) linear peptides are generally flexible molecules with entropic limitations on achieving productive biologically active conformations; (2) linear peptides are susceptible to proteolytic enzymes; and, (3) linear peptides are inherently instable. For this reason, approaches utilizing conformational and topographical constraints to restrict the number of conformational states a peptide molecule may assume have been sought. See, for example, Hruby, Life Sci., 31:189 (1982) and Hruby et al., Biochem. J. 268:249 (1990).
Head-to-tail (backbone) peptide cyclization has been used to rigidify structure and improve in vivo stability of small bioactive peptides (see Camarero and Muir, J. Am. Chem. Soc., 121:5597-5598 (1999)). An important consequence of peptide cyclization is retention of biological activity and/or the identification of new classes of pharmacological agents. Cyclic peptides have been reported that inhibit T-cell adhesion (Jois et al., J. Pept. Res., 53:18-29 (1999)), PDGF action (Brennand et al., FEBS Lett., 413:70-74 (1997)), and function as new classes of drugs (Kimura et al., (1997) J. Antibiot., 50:373-378; Eriksson et al., Exp. Cell Res., 185:86-100 (1989)).
Strategies for the preparation of circular polypeptides from linear precursors have been described. For example, a chemical cross-linking approach was used to prepare a backbone cyclized version of bovine pancreatic trypsin inhibitor (Goldenburg and Creighton, J. Mol. Biol., 165:407-413 (1983)). Other approaches include chemical (Camarero et al., Angew. Chem. Int. Ed., 37:347-349 (1998); Tam and Lu, Prot. Sci., 7:1583-1592 (1998); Camarero and Muir, Chem: Commun., 1997:1369-1370 (1997); and Zhang and Tam, J. Am. Chem. Soc. 119:2363-2370 (1997)) and enzymatic (Jackson et al., J. Am. Chem. Soc., 117:819-820 (1995)) intramolecular ligation methods which allow linear synthetic peptides to be efficiently cyclized under aqueous conditions. However, the requirement for synthetic peptide precursors has limited these chemical/enzymatic cyclization approaches to systems that are both ex vivo and limited to relatively small peptides.
One solution to this problem has been to generate circular recombinant peptides and proteins using a native chemical ligation approach. This approach utilizes inteins (internal proteins) to catalyze head-to-tail peptide and protein ligation in vivo (see, for example, Evans et al., J. Biol. Chem. 274:18359-18363 (1999); Iwai and Plückthun, FEBS Lett. 459:166-172 (1999); Wood et al., Nature Biotechnology 17:889-892 (1999); Camarero and Muir, J. Am. Chem. Soc. 121:5597-5598 (1999); and Scott et al., Proc. Natl. Acad. Sci. USA 96:13638-13643 (1999)).
Inteins are self-splicing proteins that occur as in-frame insertions in specific host proteins. In a self-splicing reaction, inteins excise themselves from a precursor protein, while the flanking regions, the exteins, become joined to restore host gene function. Inteins can also catalyze a trans-ligation self-splicing reaction. Approaches making use of the trans ligation reaction include splitting the intein into two parts and reassembling the two parts in vitro, each fused to a different extein (Southworth et al., EMBO J. 17:918-926 (1998)). A somewhat different approach uses an intein domain, and the reaction is then triggered with a thiolate nucleophile, such as DTT (Xu et al., Protein Sci. 7:2256-2264 (1998)).
The ability to construct intein fusions to proteins of interest has found several applications. For example, inteins can be used in conjunction with an affinity group to purify a desired protein (Wood et al., Nature Biotechnology, 17:889-892 (1999)). Circular recombinant fusion proteins have been created by cloning into a commercially available intein expression system (Camarero and Muir, J. Am. Chem. Soc. 121:5597-5598 (1999); Iwai and Plückthun, FEBS Lett. 459:166-172 (1999); and Evans et al., J. Biol. Chem. 274:18359-18363 (1999)). In another approach, a mechanism for in vivo split intein-mediated circular ligation of peptides and proteins via permutation of the order of elements in the fusion protein precursor has been used to express cyclic products in bacteria (Scott et al., Proc. Natl. Acad. Sci. USA, 96:13638-13643 (1999)).
Cyclic peptide libraries have been generated in phage (Koivunen et al., Biotechnology 13:265-70 (1995)) and by using the backbone cyclic proteinomimetic approach (Friedler et al., Biochemistry, 37:5616-22 (1998)). Methods for modifying inteins for the purpose of creating cyclic peptides and/or proteins have been recently described (Benzoic, et al., WO 00/36093). It is an object of this invention to utilize intein function, derived from wild-type or mutant intein structures, to generate cyclic peptide libraries in vivo. The utilization of mutant intein structures for this purpose are of particular focus since these have been optimized for function in the specific context of an intein scaffold engineered to result in peptide/protein cyclization. Methods are described for generating, identifying, and utilizing mutants with altered splicing/cyclization activity for use with cyclic peptide/protein libraries. Intein-generated cyclic libraries are described for the identification of cyclic peptides/proteins capable of altering a given cellular phenotype. Accordingly, it is an object of the invention to provide compositions and methods useful in the generation of random fusion polypeptide libraries in vivo and the use of such peptides.